Does Food Aid Stabilize Food Availability?
Christopher B. Barrett
Associate Professor
Department of Agricultural, Resource and Managerial Economics
351 Warren Hall
Cornell University
Ithaca, NY 14853-7801
Tel: 607-255-4489
Fax: 607-255-9984
Email: cbb2@cornell.edu
Http://www.cals.cornell.edu/dept/arme/staff/cbb2/
August 1999 revised version
I thank Rob Paarlberg, Tom Reardon, Terry Roe, T.N. Srinivasan, seminar audiences at Cornell and
North Carolina State, and participants at the June, 1998, International Agricultural Trade Research
Consortium and the Center for International Food and Agricultural Policy symposium on “Policy
Reform, Market Stability, and Food Security,” held in Alexandria, VA, for helpful comments. Shane
Sherlund provided excellent research assistance, and Ray Nightingale and Joel Greene graciously
made data available. Any remaining errors are mine alone.
© Copyright 1999 by Christopher B. Barrett. All rights reserved. Readers may make verbatim copies of this document for non-commercial purposes by
any means, provided that this copyright notice appears on all such copies.
Does Food Aid Stabilize Food Availability?
Abstract:
This paper explores the empirical relationship between U.S. food aid flows per capita
and nonconcessional food availability per capita in PL480 recipient economies. The
evidence suggests PL480, while perhaps modestly progressive in its distribution, fails
to stabilize food availability in recipient economies. Both increased domestic food
production – i.e., agricultural development – and commercial trade appear more
effective than food aid in increasing and stabilizing food availability per capita in lowincome economies.
JEL Codes:
Q1, O1, F1
1
Does Food Aid Stabilize Food Availability?
Food production is notoriously volatile, especially in low-income economies relatively
dependent on rainfed agriculture. Given reasonably stable per capita consumption requirements and
little interannual grain inventories carryover in poor countries, fluctuations in domestic per capita
production lead to highly variable annual import volume requirements in food importing nations.
Trade is the principal means for international food distribution at the macro level. But poorer
countries often lack the foreign exchange necessary to purchase commercially all the food needed to
meet their population’s nutritional requirements. Food aid is therefore often seen as a way to cope
with variable food import requirements and restricted commercial import capacity in low-income
economies.
While aggregate food availability is insufficient to ensure either access to or proper utilization
of nutrients to achieve food security (Sen 1981, Barrett forthcoming), aggregate availability is
nonetheless a necessary condition for food security. Food insecurity is inevitable within an economy
lacking enough food to satisfy all its population’s nutritional needs, even if distributed perfectly
equitably and without loss to spoilage or waste. Ensuring adequate aggregate food availability has
been, and remains today, a serious challenge in much of the low-income world. Average per capita
daily energy and protein availability of 2244 kilocalories and 54.9 grams, respectively, 1961-95 in the
low-income economies fell below international recommended nutrient intake levels ( WHO 1985,
FAO 1999). Even today, a majority of the low-income countries have per capita daily energy
availability of less than 2500 kilocalories, signaling that availability remains an issue in advancing
2
universal access to sufficient and appropriate food.
The basic logic of food aid for food security is therefore simple. In so far as food aid is meant
to address food availability shortfalls that might cause undernutrition, food aid should flow in
response to such shortfalls.1 This raises the question of how one defines a food availability shortfall.
In this paper I use each of two reasonable alternatives. A shortfall in cross-section reflects scarcity
relative to others. A shortfall in time series reflects scarcity relative to trend availability. Food aid
for security should therefore flow disproportionately to countries exhibiting low per capita
nonconcessional food availability (NA) — a cross-sectional shortfall — a sharp negative deviation
from trend NA — a time series shortfall — or both. But food aid in fact flow to recipient economies
in such a manner? That is the question tackled in this paper, as I explore the empirical relationship
between food aid flows per capita from the United States’ PL480 programs and nonconcessional food
availability per capita in PL480 recipient economies. If food aid indeed stabilizes food availability,
then per capita food aid flows should be inversely related to recipients’ per capita nonconcessional
food availability, in terms of levels, deviations from trend, or both. This is an empirically testable
hypothesis that, to the best of my knowledge, has not yet been studied.
Nonconcessional Food Availability Trends in PL480 Recipient Economies
Let me begin with some definition of terms and data description. Because individual
physiology drives nutritional needs, and in order to be able to compare countries with vastly different
human populations, all figures reported are in per capita terms. In order to work with readily
comparable series without introducing serious aggregation bias problems, I use cereals volumes to
proxy total food production, nonconcessional availability (production plus commercial imports), and
3
aid flows per capita.2 Annual production, commercial import, and population data, 1961-95, were
provided by the Food and Agriculture Organization of the United Nations, while disaggregated (by
year, commodity, Title, and recipient country) PL480 food aid flows data were obtain from the U.S.
Department of Agriculture’s Economic Research Service. The data cover 124 different recipient
economies, representing all PL480 recipients during the period other than Japan and developed
European economies.3
For those countries that achieved independence after 1961, only
independence-era data are used, yielding an uneven panel of data.
The food available to feed a country’s residents comes from one of four sources: domestic
production, domestic inventories, commercial imports from abroad, or food aid inflows from abroad.
This paper looks at how the latter source, food aid, covaries with the first three — which together
make up the category nonconcessional food availability (NA) — in order to establish whether food
aid helps stabilize aggregate food availability. A data problem emerges immediately. Reliable cereals
inventories data are unavailable for most countries, particularly poorer food aid recipients. But since
interannual cereals stocks per capita are generally quite small in developing countries,4 the unrealistic
limiting assumption used here — that per capita inventories equal zero — probably has little effect
on the forthcoming analysis. I should also point out that I do not include total food aid flows from
all donors; the analysis considers only PL480 shipments from the United States. But since PL480
comprised about two-thirds of global food aid, 1961-95, the data used here should capture the basic
patterns prevailing more broadly.
Own production and commercial trade account for the vast majority of cereals availability in
PL480 recipient nations. Pooling across years and recipients, domestic production’s mean (median)
proportion of aggregate national cereals availability, defined as production plus commercial imports
4
plus PL480 receipts, was 69.3 (80.2) percent.5 Mean (median) commercial imports accounted for
another 28.6 (17.6) percent of recipient country food availability, leaving only a tiny fraction covered
by PL480 shipments most years in most recipient countries, as can be seen in Table 1 and graphically
in Figure 1. Given that PL480 flows rarely comprise more than a negligible proportion of total food
availability in recipient countries, this suggests that food aid can play, at best, a very limited stabilizing
role. The meagerness of food aid is only one reason for its limited efficiency, however. Even modest
amounts of food aid could have a significant effect if targeted and timed well.
The 1961-95 Green Revolution era of rapid biochemical improvements to cropping systems,
brought unprecedentedly rapid annual average growth of 0.5 percent in global cereals production per
capita (Barrett forthcoming). PL480 recipients, however, lagged significantly behind. Annual average
growth rates in production and NA for each PL480 recipient, 1961-95, were estimated by equations
(1) and (2), respectively.
ln(PRODUCTIONt)="0P+"1P YEAR+,Pt
(1)
ln(NAt)="0N+"1N YEAR+,Nt
(2)
Across the 124 PL480 recipients, the median annual growth rate in per capita cereals production was
-0.2 percent, i.e., more than half (53%) the countries suffered negative average annual growth. The
voluminous literature on food aid emphasizes its potential disincentive effects on recipient country
production, and perhaps the sluggish growth in recipient production reflects this (Maxwell and Singer
1979; Ruttan 1993; Barrett forthcoming).
Rapid growth in PL480 recipients’ commercial cereals imports has made up for sluggish
cereals production growth. The median annual growth rate in per capita nonconcessional cereals
availability was 0.5 percent, the same as the global growth rate in per capita cereals production (and
5
therefore global NA). Still, more than one-third (37%) of the countries exhibit negative average
annual growth even in NA.
While the trends are informative, the variability around trend NA is of at least as much
interest, in that this reflects short-run instability in food supplies to which food aid is supposed to at
least partly respond if it is to serve food security objectives. The estimated standard errors of the
residuals of equations (1) and (2), e8
Pt
and e8Nt ,6 capture this interannual variability around trend
production and nonconcessional availability, respectively. In the next section, I study the empirical
relationship between PL480 flows and e8Nt in order to test whether PL480 flows stabilize food
availability (i.e., covary negatively with shocks to trend nonconcessional food availability). But first,
let’s quickly look more carefully at the regression results from equations (1) and (2).
Among PL480 recipients there exists a negative univariate relationship between the average
annual growth rate and the standard deviation around trend cereals production per capita. Let vP be
the standard deviation of the e8Pt series and vN be the standard deviation of the e8Nt series. Regressing
vP on a81P and an intercept term yields a coefficient estimate of -0.914 (with a standard error of the
estimate of 0.465).7 This crude result supports the intuitive hypothesis that faster growth in cereals
productivity tends to bring with it greater stability around trend per capita production. Put
differently, agricultural development appears important not only to increasing developing countries’
food availability but also to stabilizing food availability.
Moreover, because production makes up the bulk of countries’ food availability (Figure 1),
domestic food production drives nonconcessional food availability. The simple ordinary least squares
regression of the annual average growth rates in PL480 recipients’ nonconcessional cereals
availability, a81N , on production per capita, a81P , shows that the two are positively and statistically
6
significantly related, as one would expect.8 The statistically significant, sub-unit (0.644) estimated
coefficient also reflects the effective role that commercial international trade plays in stabilizing food
availability in developing countries. NA responds at less than a one-for-one rate to changes in
domestic production. Commercial trade’s stabilizing effect is also reflected by the fact that vN<vP in
more than 80 percent of the sample recipients. The mean reduction in the standard deviation of per
capita cereals volumes is greater than eight percent per annum, from v2 P = 0.237 to v2 N = 0.156.
Commercial food trade contributes significantly to the stabilization of food availability in developing
countries.
While commercial cereals trade plays a crucial role in stabilizing food availability in low- and
middle-income countries, binding foreign exchange constraints nonetheless commonly limit the
capacity of poorer countries to dampen food supply volatility through commercial markets. At 15.6
percent, the standard deviation of NA per capita in PL480 recipients remains more than three times
the world standard deviation around trend of 4.7 percent. Indeed, 122 of 124 PL480 recipients
evince more variable NA than the global rate (all except Georgia and Russia). Given the residual
need for food consumption smoothing in developing countries, the core question remains: have
PL480 food aid shipments helped to stabilize food availability in the face of extraordinary variability
in recipients’ nonconcessional food availability? Put differently, how effectively has PL480 targeted
food insufficiency at the national level?
PL480 Responsiveness To Need: An Empirical Analysis
PL480 flows have dominated global food aid since the program’s inception in 1954. At risk
of some oversimplification, there are two basic types of PL480 food aid: program (Titles I and III)
7
and emergency (Title II). The operational distinction between them is perhaps best reflected by the
division of responsibility over PL480 between executive branch agencies. Title II PL480 distributions
are directed by the U.S. Agency for International Development with an expressed objective of
development and relief. Other PL480 distributions are handled by the U.S. Department of
Agriculture, which also aims to promote U.S. food exports, and in the era of generous crop price
support programs used to use food aid to dispose of considerable government-held food inventories.
A primary reason to examine PL480 flows disaggregated between program and emergency assistance
is the popular belief that Title II flows are more responsive to need, particularly to short-term
instability in recipient country NA. Yet program food aid has long dominated PL480 flows. Between
1954 and 1995, Titles I and III of PL480 accounted for better than 80 percent of the more than 300
million metric tons of U.S. food aid and more than half of total worldwide food aid flows. That said,
program (emergency) food aid has steadily diminished (grown) in importance over the past twenty
years. Program flows averaged 86% of PL480 deliveries and were at least 80% each year prior to
1973, but averaged only 72%, 1973-95, and were above 80% only 3 of those 23 years. Title II
shipments surpassed Title I flows for the first time only in 1993.
There are at least five interrelated reasons to be skeptical about the effectiveness of PL480
food aid in dampening variability in recipient country food availability. First, previous studies have
shown US food aid has been driven largely by considerations other than food security, with relatively
little targeting toward countries with pronounced food deficits (Ruttan 1993, 1995; Ball and Johnson
1996, Barrett forthcoming). Surplus disposal and trade promotion objectives and especially
geopolitical considerations have largely dominated food aid’s history. Political objectives tend to
trump food security concerns in Washington. Second, and related to the first, PL480 flows have
8
shown far greater persistence over the years than is consistent with the claim that they respond to
transitory nonconcessional food availability shortfalls in recipient countries (Barrett 1998, Barrett et
al. 1999). Third, PL480 flows — indeed bilateral flows more generally — have proved procyclical
in aggregate, not countercyclical, because they are budgeted in monetary rather than volume terms
(Barrett forthcoming). When world market prices are high, recipient country commercial imports fall
and food aid needs grow, but food aid volumes also fall because the budget covers only a smaller
volume at a higher price. Fourth, food aid is not fully additional, meaning that food aid receipts
consistently replace 60-80 percent of the commercial food imports recipient economies would have
made (Von Braun and Huddleston 1988, Barrett forthcoming). Such fungibility necessarily limits the
efficacy of food aid in stabilizing food availability. Fifth, until quite recently few good early warning
systems existed to anticipate emergencies accurately, so food aid deliveries are largely reactive and
therefore often ill-timed. Of these five concerns, only the latter situation may be improving
significantly in the case of PL480, although early warning systems continue to have a spotty
performance record (Barrett forthcoming).
The simplest way to establish whether food aid dampens the variability of recipient country
food availability is to estimate the empirical relationship between food aid flows per capita, FA, and
both the levels, NA, and the deviations from trend NA , e8N , from equation (2). If food aid flows to
those most in absolute need, as reflected by a negative correlation between PL480 and NA levels,
then food aid can be described as progressive. If food aid responds negatively to deviations from
national trend NA, then FA has a stabilizing, countercyclical effect. The magnitude of the latter
relationship is of particular interest as it indicates the compensation proportion, i.e., the proportion
of a shortfall that is made up for by PL480 flows.
9
Since FA is a nonnegative variable often taking zero value, this relationship is estimated by
the Tobit model:
FAit = $0 +$1 e8Nit + $2 NAit + Tit
if FAit > 0
(3a)
FAit = 0
if FAit = 0
(3b)
where i indexes recipient countries and t indexes years. $1 captures the stabilization effect of food
aid, while $2 reflects the distributional effect. Since the data are pooled cross-sectional and time
series, it is necessary to test first for fixed effects in cross-section, intertemporally, or both. The
specification test statistics suggest it is necessary only to control for unobserved region-specific
effects.9 A bit later, I consider the results of country- and year-specific estimation of (3) to see
whether imposing a universal relationship masks different relations in a nontrivial subsample of
countries (it doesn’t).
Several interesting results appear in Table 2. The b8 1 and b8 2 estimates are of uniformly low
magnitude, most of the b8 1 (b8 2) estimates are positive (negative), and most of the estimates are not
statistically significantly different from zero. The low magnitudes reflect in part the negligible
contribution of food aid to aggregate food availability in food recipient economies, as suggested
earlier by Figure 1. Since the b8 1 coefficient estimates represent compensation proportions the
negative and statistically significant b8 2 estimate in the full sample suggests that PL480 has flowed
somewhat more to food scarce than food abundant economies, although the associated elasticity,
estimated at sample means is only -0.04. There appears to be only very modest global progressivity
to PL480 distribution. But the counterintuitively positive signs of the b8 1 estimates suggest that food
aid flows have been, if anything, procyclical, not countercyclical on average. In the full sample, one
cannot reject at any reasonable level of statistical significance the null hypothesis that PL480 flows
10
are uncorrelated with deviations from recipients’ trend per capita food availability. So while the data
weakly support the claim that PL480 has been (modestly) distributionally progressive, they in no way
support the claim that PL480 has stabilized food availability in recipient economies.
These results hold not only in the full pool of 124 developing country PL480 recipients, but
also in three subsamples of particular interest. In the 1960s and into the 1970s food aid — especially
program (Title I) PL480 — was disproportionately concentrated on South Asia. For South Asia,
home to the largest number of the world’s food insecure, PL480 flows have been statistically
significantly procyclical while the estimated progressivity effect is not statistically significantly
different from zero. Since the world food crisis of the mid-1970s, PL480 — especially humanitarian
(Title II) flows — have been disproportionately focused on Sub-Saharan Africa (SSA), the only
world region in which the proportion of the population suffering food insecurity has not fallen
significantly for a generation. PL480 flows to SSA are of particularly low magnitude and statistical
significance, and of the wrong (positive) sign to support either the claim that PL480 has stabilized
African food availability or the claim that food aid has flowed most generously to those countries
most in need. Finally, I also ran the regression for an international group of countries whose PL480
programs (or termination of those programs) are widely recognized as geopolitically motivated. One
might suspect that the estimation results from the full sample are contaminated by the inclusion of
countries whose PL480 programs have been plainly driven by non-economic and non-humanitarian
considerations. The curious result is that while the magnitudes and statistical significance of the
parameter estimates are also low, only in this subsample do we get negative point estimates for both
b8 1 and b8 2 . So the subset of geopolitically motivated PL480 country programs do not seem to distort
the estimation results in the full sample.10
11
The results are also qualitatively unchanged when we reestimate off emergency (Title II) food
aid alone or program (Titles I and III) food aid alone, as shown in Tables 3 and 4, respectively. As
shown in the rightmost column of Table 3, only in the case of Title II PL480 to the set of
geopolitically motivated recipients does food aid have both stabilization and distributional effects that
are statistically significantly negative. PL480 food aid, of any sort, has not stabilized food availability
on average in recipient economies, even though its distribution has been modestly progressive on a
global -- if not always regional -- scale. So the widespread claim that humanitarian (i.e., Title II) food
aid is somehow more responsive to need finds no support in the country-level data, due likely to the
factors enumerated earlier.
Given the idiosyncracies of PL480 programs in individual recipient countries, and the
evolving rhetoric and operational codes of PL480 over 35 years, one might be justifiably skeptical
of the results from regressions using data pooled across countries and years. The same qualitative
results obtain, however, when one examines the distribution of country- or year-specific estimation
results.11 For example, the distribution of country-specific estimates of model (3) shows that most
parameter estimates are statistically insignificantly different from zero, extraordinarily few b8 1 estimates
are less than -0.1 (which would imply ten percent average compensation effect from PL480 flows)
or even statistically significantly negative, and PL480 most commonly flows procyclically around
recipients’ food availability trend, not countercyclically (Table 5). The consistency between the
patterns found in the distribution of parameter estimates derived from the country-specific time series
and the estimated from the pooled sample reported in Tables 2-4 suggests that country-specific
differences due to variation in local PL480 operations or recipient country policy do not explain the
failure of food aid to stabilize national food availability.
12
Two country examples illustrate how the inefficacy of PL480 in stabilizing food availability
arises not just from the small volume of aid flows but also from systematic mistiming (Figure 2).
Ethiopia is currently and historically the leading food aid recipient in sub-Saharan Africa. The 1984
famine there drew unprecedented international attention. But PL480 deliveries increased only
modestly in 1984 when nonconcessional food availability plummeted. Rather, food aid shipments
boomed in 1985 and 1986, when recovery was already well underway. Indeed, the all-time high for
per capita PL480 deliveries to Ethiopia was 1986, which was also the second most plentiful year of
nonconcessional food per capita in a fifteen year span in Ethiopia! At the national level at least, wellintentioned PL480 shipments arrived when it was least needed. Similarly, Peru is the only country
to receive PL480 flows every year since the program’s inception in 1954. In 1994, US food aid flows
to Peru more than doubled although nonconcessional food availability in Peru also jumped almost
twenty percent that year. By contrast, during the earlier, steady decline in nonconcessional food
availability in Peru from 1987-90, PL480 flows also fell steadily. Sen (1981) famously showed that
food availability is not sufficient to ensure food security. The empirical evidence analogously
suggests decreased (increased) nonconcessional food availability is not sufficient to ensure greater
(lesser) PL480 food aid flows to maintain sufficient supplies in low-income countries.
Although not reported here, the same basic results obtain in cross-section, in the distribution
of year-specific estimates.12 Moreover, the common claim that improvements have been made to
PL480 operations based on past lessons learned finds no support in these estimates. There were only
five years during the period 1961-95 in which both the b8 1 and b8 2 point estimates were positive in
cross-section. Three of the five came in the 1990s, in emergency, program, and pooled PL480
samples alike. So the claim that PL480 distribution meets distributional and stabilization goals more
13
effectively today than in the (Cold War) past finds no support in these data.
A final, cautionary note is in order. The macro data used in this analysis cannot capture
prospective international or intertemporal variation in the efficacy of intranational food distribution
systems in reaching food insecure subpopulations. The analysis reported here necessarily stops at the
recipient’s port since the data used are national aggregates. So although these results suggest food
aid is ineffective in stabilizing food availability at the macro level, it is theoretically plausible that food
aid targeting within recipient economies is so effective that food aid nonetheless stabilizes food
availability for particular food-insecure communities, households or individuals. There is certainly
anecdotal evidence of emergency food aid distributions proving helpful in averting humanitarian
disasters on short notice (Shaw and Clay 1993).
There has also been progress in adapting the
modalities of emergency food aid delivery, although this seems more true for World Food Programme
distributions than PL480 flows (Barrett 1998, forthcoming; Clay et al. 1996). Nonetheless,
emergency food aid deliveries are often mistimed, misallocated, or both, sometimes doing more harm
than good (Jackson with Eade 1992; Stewart 1998). The only published study of which I am aware
that uses micro-level data to investigate community- and household-level food aid targeting finds that
food aid flows disproportionately to the most food secure regions and households in Ethiopia (Clay
et al. 1999).13 No systematic micro-level evidence seems to exist to demonstrate that even though
food aid is remarkably poorly targeted at macro level, it is well enough targeted at micro level to have
net positive effects in stabilizing the poor’s access to food. Given the uneven performance of PL480,
the evidence presented here puts the burden of proof on those who would claim that PL480 food aid
is effectively enough targeted intranationally to overcome its insignificant macro-level effects in
stabilizing recipient food availability.
14
Conclusions
Improving food security and health and nutritional outcomes around the world will require
dampening the extraordinary variability in per capita food availability in low-income economies.
Improved food productivity and commercial international trade appear far more useful than PL480
food aid in achieving that objective. The small volumes, opaque allocation mechanisms, and
bureaucratically cumbersome procurement procedures behind PL480 have made food aid a relatively
ineffective instrument of either stabilization or redistribution. While there are surely particular
emergencies and distribution modalities through which food aid can play an effective role in
stabilizing and improving food availability at the micro level of individual communities, households,
and individuals, commercial trade and more rapid domestic food productivity growth both appear
more effective in stabilizing developing national food availability in the regular course of
development. Perhaps if food aid were targeted entirely toward relieving food insecurity it could be
a more effective instrument. But food aid has long been intensely political, serving many masters.
So long as that remains the case, food aid is unlikely to stabilize per capita food availability
effectively.
15
References
Ball, Richard and Christopher Johnson (1996), “Political, Economic, and Humanitarian Motivations
for PL480 Food Aid: Evidence from Africa,” Economic Development and Cultural Change,
44, 3: 515-537.
Barrett, Christopher B. (1998), “Food Aid: Is It Development Assistance, Trade Promotion, Both
or Neither?” American Journal of Agricultural Economics, 80, 3: 566-571.
Barrett, Christopher B. (Forthcoming), “Food Security and Food Assistance Programs,” in Bruce
L. Gardner and Gordon C. Rausser, editors, Handbook of Agricultural Economics
(Amsterdam: Elsevier Science).
Barrett, Christopher B., Sandeep Mohapatra, and Donald L. Snyder (1999), “The Dynamic Effects
of U.S. Food Aid,” Economic Inquiry, 37, 4: 647-656.
Clay, Daniel C., Daniel Molla, and Debebe Habtewold (1999), “Food Aid Targeting in Ethiopia: A
Study of Who Needs It and Who Gets It,” Food Policy, 24, 3: 391-409.
Clay, Edward J., Sanjay Dhiri and Charlotte Benson (1996), Joint Evaluation of European Union
Programme Food Aid: Synthesis Report (London: Overseas Development Institute).
FAO, Food Balance Sheets, (online at http://apps.fao.org/).
Jackson, Tony with Deborah Eade (1992) Against The Grain: The Dilemmas of Project Food Aid
(Oxford: OXFAM).
Ruttan, Vernon W. (1993), Why Food Aid? (Baltimore: Johns Hopkins University Press).
Ruttan, Vernon W. (1995), United States Development Assistance Policy: The Domestic Politics of
Foreign Economic Assistance (Baltimore: Johns Hopkins University Press).
Sen, Amartya (1981), Poverty and Famines (Oxford: Oxford University Press).
Shaw, John and Edward Clay, eds. (1993), World Food Aid: Experiences of Recipients and Donors
(Portsmouth, NH: Heinemann).
Stewart, Frances (1998), “Food Aid During Conflict: Can One Reconcile Its Humanitarian,
Economic and Political Economy Effects?” American Journal of Agricultural Economics,
80, 3: in press.
Von Braun, Joachim, and Barbara Huddleston (1988), “Implications of Food Aid for Price Policy
16
in Recipient Countries.” Agricultural Price Policy for Developing Countries, edited by John
W. Mellor and Raisuddin Ahmed. Baltimore: Johns Hopkins University Press.
World Health Organization (1985), Energy and Protein Requirements: Report of a Joint
FAO/WHO/UNU expert consultation, Technical Report Series no. 724.
17
Table 1: Shares of Aggregate Cereals Availability
PL480 Recipients, 1961-95
Own Production
Commercial Imports
PL480
Mean
0.693
0.286
0.021
Median
0.803
0.176
0.002
Std. Deviation
0.294
0.286
0.047
Maximum
1.000
1.000
0.644
Minimum
0
0
0
18
Table 2: Tobit Regression Results, All PL480 (Titles I, II, and III)
All 124 Countries
South Asia
Sub-Saharan Africa
Geopolitically
Motivated
$1
(stabilization effect)
0.001
(0.004)
0.038
(0.012)
0.001
(0.002)
-0.008
(0.010)
$2
(distributional effect)
-0.029
(0.010)
-0.029
(0.037)
0.014
(0.012)
-0.008
(0.029)
ln(L)
-553.0
-231.6
-319.6
-188.0
n
3838
210
1453
880
Standard errors in parentheses.
Tobit regressions including regional dummy variables to control for fixed effects. Regions included are Central
Africa, Central America, East Africa, East Asia, Europe, Middle East, North Africa, North America, South
America, South Asia, Southeast Asia, Southern Africa, former USSR, West Africa, West Asia, and former
Yugoslavia. South America is the base for the global model, West Africa is the base for the Sub-Saharan Africa
model, and Europe is the base for the geopolitically motivated model. No fixed effects were found in the South
Asia model.
South Asia: Afghanistan, Bangladesh, India, Nepal, Pakistan, Sri Lanka.
Sub-Saharan Africa: Angola, Benin, Botswana, Burkina Faso, Burundi, Cameroon, Canary Islands, Cape Verde,
Central African Republic, Chad, Comoros, Congo, Côte d’Ivoire, Djibouti, Equatorial Guinea, Eritrea, Ethiopia,
Gabon, Gambia, Ghana, Guinea, Guinea-Bissau, Kenya, Lesotho, Liberia, Madagascar, Malawi, Mali,
Mauritania, Mauritius, Mozambique, Niger, Nigeria, Rwanda, Sao Tome Principe, Senegal, Seychelles, Sierra
Leone, Somalia, Sudan, Swaziland, Tanzania, Togo, Uganda, Zaire, Zambia, Zimbabwe.
Geopolitically Motivated: Afghanistan, Belarus, Bosnia, Cyprus, Dominican Republic, Egypt, El Salvador,
Ethiopia, Gaza Strip, Haiti, Honduras, Iran, Iraq, Israel, Jordan, Korea, Laos, Lebanon, Nicaragua, Philippines,
Poland, Russia, Somalia, Sudan, Syria, Taiwan, Ukraine, Vietnam, Zaire.
19
Table 3: Tobit Regression Results, Emergency (Title II) PL480
All 124 Countries
South Asia
Sub-Saharan Africa
Geopolitically
Motivated
$1
(stabilization effect)
0.001
(0.001)
0.002
(0.002)
0.001
(0.001)
-0.00003
(0.001)
$2
(distributional effect)
-0.006
(0.001)
-0.008
(0.007)
0.005
(0.004)
-0.009
(0.004)
ln(L)
-2299.2
-385.4
-619.4
-823.7
3838
210
1453
880
n
Standard errors in parentheses. Same notes apply as on Table 2.
Table 4: Tobit Regression Results, Program (Titles I and III) PL480
All 124 Countries
South Asia
Sub-Saharan Africa
Geopolitically
Motivated
$1
(stabilization effect)
0.0004
(0.004)
0.037
(0.011)
0.001
(0.002)
-0.008
(0.009)
$2
(distributional effect)
-0.025
(0.009)
-0.023
(0.035)
0.009
(0.009)
-0.003
(0.027)
ln(L)
-623.8
-240.9
-374.7
-206.3
n
3838
210
1453
880
Standard errors in parentheses. Same notes apply as on Table 2.
20
Table 5: Descriptive Statistics of Country-Specific Tobit Regression Results
All PL480
Title II only
Titles I and III only
Stabilization effects:
$1 >0 (%)
67.7
67.6
71.8
Reject H0: $1=0 (%)
26.6
28.8
43.5
6.4
8.9
9.9
10th percentile $1
-0.084
-0.002
-0.074
Median $1
0.014
0.004
0.008
90th percentile $1
0.152
0.055
0.345
Distributional effects:
$2 >0 (%)
36.3
30.6
56.4
Reject H0: $2=0 (%)
17.7
17.1
34.6
11.3
11.7
24.3
10th percentile $2
-0.091
-0.034
-0.187
Median $2
-0.008
-0.003
-0.012
90th percentile $2
0.074
0.017
0.088
124
111
78
o/w $1 <0 (%)
o/w $2 <0 (%)
n
21
Figure 1
Cumulative Frequency
Aggregate Cereals Availability Shares
PL 480 Recipients, 1961-95
1.0
0.8
0.6
0.4
0.2
0.0
0.0
0.2
PL480
0.4
0.6
Proportion
Production
0.8
Comm. Imports
1.0
22
Figure 2: Comovement in PL480 and Nonconcessional Food Availability
23
Endnotes
1. Since food aid has traditionally been a macro or sectoral policy instrument, distributed mainly
on a government-to-government basis, its relationship to aggregate food availability is an
important question, albeit not the only important question surrounding food aid. For broad
surveys, see especially Ruttan (1993), Shaw and Clay (1993), and Barrett (forthcoming).
2. Cereals accounted for more than 60 percent of calories and protein in low-income countries,
1961-95, and are by far the single largest source of macronutrients in all low and middle-income
countries today (FAO 1999).
3. The only 1961-95 PL480 recipients omitted from the data set are Austria, Finland, France,
Germany, Hungary, Iceland, Italy, Japan, Malta, and Spain.
4. While on-farm storage for autoconsumption is considerable, if largely unmeasured, the limited
available evidence indicates the vast majority of these stocks are consumed within the year (Sahn
1989).
5. Note that this “aggregate” cereals availability figures omits both food aid receipts other than
PL480 shipments and domestic cereals inventories, although these are both relatively small
volumes.
6. I distinguish unbiased, consistent regression estimates from the true but unknown population
parameters by using Roman rather than Greek letters and the caret (8) symbol.
7. Unlike, cereals production, NA variability and growth rates are unrelated in the set of PL480
recipient economies.
8. The OLS regression result is: a81N = -0.006 + 0.644 a81P
(0.102)
9. Using the general model form FAit=$0 +$1 e8it +$2 NAit +3j *j REGIONjit +3t (t YRit + Tit if
FAit > 0, likelihood ratio tests of the joint restrictions *j =0 œ j, (t =0 œ t, or both yield test
statistics that uniformly support rejecting the null hypothesis of *j =0 œ j at any level of statistical
significance for program, emergency, or all PL480 aid, and uniformly fail to support rejecting the
null hypothesis of (t =0 œ t at even the ten percent significance level for program, emergency, or
all PL480 flows. Test details are available from the author by request.
10. The qualitative results in the rightmost column of Table 2 are robust to each of the several
combinations of countries tried in the “geopolitically motivated” set.
11. In estimating the country-specific time series, the regression residuals were subjected to
diagnostic portmanteau statistics for autocorrelation. In those instances where autocorrelation
was evident, appropriate correction was made using Box-Jenkins techniques.
24
12. A table presenting these results is available from the author by request.
13. Note that cross-sectional studies like Clay et al. (1999) test only what I term the
“progressivity” of food aid distribution. No one appears to have yet studied the dynamic
“stabilization” effects at the micro level.